Re: [PR] [SVE] Add support for scalable data type strings [tvm]

ekalda merged PR #16612: URL: https://github.com/apache/tvm/pull/16612 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail:

(tvm) branch main updated: [SVE] Add support for scalable data type strings (#16612)

This is an automated email from the ASF dual-hosted git repository. ekalda pushed a commit to branch main in repository https://gitbox.apache.org/repos/asf/tvm.git The following commit(s) were added to refs/heads/main by this push: new 563ef9587c [SVE] Add support for scalable data type

Re: [PR] [SVE] Add support for scalable data type strings [tvm]

ekalda commented on PR #16612: URL: https://github.com/apache/tvm/pull/16612#issuecomment-1966121711 Thanks @lhutton1! -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To

Re: [PR] [BugFix] Fix Crash Cases Caused by "__tvm_meta__ = None" [tvm]

Hzfengsy commented on PR #16634: URL: https://github.com/apache/tvm/pull/16634#issuecomment-1966380114 cc @tqchen -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To

Re: [PR] [SVE] Add support for scalable data type strings [tvm]

lhutton1 commented on PR #16612: URL: https://github.com/apache/tvm/pull/16612#issuecomment-1966568621 Apologies for missing this previously, I've uploaded #16649 which I'd expect to fix the failing test case -- This is an automated message from the Apache Git Service. To respond to the

Re: [PR] [SVE] Add support for scalable data type strings [tvm]

ekalda commented on PR #16612: URL: https://github.com/apache/tvm/pull/16612#issuecomment-1966587228 Any idea why the CI didn't catch this? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the

Re: [PR] [SVE] Add support for scalable data type strings [tvm]

tqchen commented on PR #16612: URL: https://github.com/apache/tvm/pull/16612#issuecomment-1966508091 there is a regression due to concurrent merge, cc @Lunderberg -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the

Re: [PR] [SVE] Add support for scalable data type strings [tvm]

lhutton1 commented on PR #16612: URL: https://github.com/apache/tvm/pull/16612#issuecomment-1966532455 thanks for raising @tqchen, I'm taking a look now -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to

Re: [PR] [SVE] Add support for scalable data type strings [tvm]

lhutton1 commented on PR #16612: URL: https://github.com/apache/tvm/pull/16612#issuecomment-1966648171 CI was last run on the patch yesterday (26th Feb) while the PR that caught the issue was merged a week ago https://github.com/apache/tvm/pull/16563. I didn't rebase before pushing, but

Re: [PR] [Relax] Allow R.Prim('bool') in relax::If and assert_op [tvm]

Lunderberg commented on code in PR #16642: URL: https://github.com/apache/tvm/pull/16642#discussion_r1504405971 ## python/tvm/relax/transform/transform.py: ## @@ -463,6 +463,16 @@ def KillAfterLastUse() -> tvm.ir.transform.Pass: return _ffi_api.KillAfterLastUse() # type:

Re: [PR] [Relax] Allow R.Prim('bool') in relax::If and assert_op [tvm]

Lunderberg commented on code in PR #16642: URL: https://github.com/apache/tvm/pull/16642#discussion_r1504424495 ## src/relax/transform/compute_prim_value.cc: ## @@ -0,0 +1,94 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license

Re: [PR] [TVMScript] Allow use of relax.Expr with void type as a statement [tvm]

slyubomirsky commented on code in PR #16641: URL: https://github.com/apache/tvm/pull/16641#discussion_r1504846413 ## python/tvm/script/parser/relax/parser.py: ## @@ -274,7 +274,21 @@ def post_visit_local_function(self: Parser, node: doc.Expr) -> None:

Re: [PR] [Relax] Allow R.Prim('bool') in relax::If and assert_op [tvm]

Lunderberg commented on code in PR #16642: URL: https://github.com/apache/tvm/pull/16642#discussion_r1504921950 ## python/tvm/relax/transform/transform.py: ## @@ -463,6 +463,16 @@ def KillAfterLastUse() -> tvm.ir.transform.Pass: return _ffi_api.KillAfterLastUse() # type:

Re: [PR] [BugFix] Fix Crash Cases Caused by "__tvm_meta__ = None" [tvm]

slyubomirsky commented on PR #16634: URL: https://github.com/apache/tvm/pull/16634#issuecomment-1967667539 I had this issue come up in https://github.com/apache/tvm/pull/16569. I think these tests probably were not being executed before. I am also pretty sure it's safe to just remove those

Re: [PR] [Relax] Allow R.Prim('bool') in relax::If and assert_op [tvm]

slyubomirsky commented on code in PR #16642: URL: https://github.com/apache/tvm/pull/16642#discussion_r1504800879 ## src/relax/transform/compute_prim_value.cc: ## @@ -0,0 +1,94 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license

Re: [PR] [Relax] Allow R.Prim('bool') in relax::If and assert_op [tvm]

slyubomirsky commented on code in PR #16642: URL: https://github.com/apache/tvm/pull/16642#discussion_r1504800473 ## python/tvm/relax/transform/transform.py: ## @@ -463,6 +463,16 @@ def KillAfterLastUse() -> tvm.ir.transform.Pass: return _ffi_api.KillAfterLastUse() #

(tvm) branch main updated (563ef9587c -> ff3716b83a)

This is an automated email from the ASF dual-hosted git repository. tqchen pushed a change to branch main in repository https://gitbox.apache.org/repos/asf/tvm.git from 563ef9587c [SVE] Add support for scalable data type strings (#16612) add ff3716b83a [TVMScript] Represent

Re: [PR] [Relax] Allow R.Prim('bool') in relax::If and assert_op [tvm]

slyubomirsky commented on code in PR #16642: URL: https://github.com/apache/tvm/pull/16642#discussion_r1505049818 ## python/tvm/relax/transform/transform.py: ## @@ -463,6 +463,16 @@ def KillAfterLastUse() -> tvm.ir.transform.Pass: return _ffi_api.KillAfterLastUse() #

Re: [PR] [Relax] Allow R.Prim('bool') in relax::If and assert_op [tvm]

slyubomirsky commented on code in PR #16642: URL: https://github.com/apache/tvm/pull/16642#discussion_r1505049818 ## python/tvm/relax/transform/transform.py: ## @@ -463,6 +463,16 @@ def KillAfterLastUse() -> tvm.ir.transform.Pass: return _ffi_api.KillAfterLastUse() #

Re: [I] [Tracking Issue] [WebGPU] Supporting DP4A in WebGPU backend [tvm]

tqchen commented on issue #16627: URL: https://github.com/apache/tvm/issues/16627#issuecomment-1967465410 Thanks @Jiawei-Shao ! contribution is more than welcomed -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the

Re: [PR] [TVMScript] Allow use of relax.Expr with void type as a statement [tvm]

Lunderberg commented on PR #16641: URL: https://github.com/apache/tvm/pull/16641#issuecomment-1967668483 True. I'd see the mismatch as the degree to which relax should be compatible with the python environment. For lisp, I'd expect `nil` to be the empty tuple, but for Python, I'd expect

Re: [PR] [Arith] Provide tighter ConstIntBounds for special cases [tvm]

tqchen commented on PR #16588: URL: https://github.com/apache/tvm/pull/16588#issuecomment-1967461907 Thank you @Lunderberg ! -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment.

Re: [PR] [TVMScript] Represent tir::builtin::ret() using python "return" [tvm]

tqchen merged PR #16640: URL: https://github.com/apache/tvm/pull/16640 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail:

Re: [PR] [Unity][Parser] Check well-formedness in the parser [tvm]

slyubomirsky commented on PR #16569: URL: https://github.com/apache/tvm/pull/16569#issuecomment-1967569915 Note that `tests/python/tir-transform/test_tir_transform_hoist_if.py::test_hoisting_block_scope_4` and `test_tir_transform_force_narrow_index_to_i32.py::test_thread_axis2` also fail

Re: [PR] [BugFix] Fix Crash Cases Caused by "__tvm_meta__ = None" [tvm]

slyubomirsky commented on PR #16634: URL: https://github.com/apache/tvm/pull/16634#issuecomment-1967669197 > Run these test cases in local with this patch, below 3 test cases still will be failed because of TIR structure isn't equal, @Hzfengsy @tqchen can you help to see them? Thanks. >

Re: [PR] [Unity][Parser] Check well-formedness in the parser [tvm]

slyubomirsky commented on code in PR #16569: URL: https://github.com/apache/tvm/pull/16569#discussion_r1505317426 ## tests/python/tir-base/test_tir_specialize.py: ## @@ -275,7 +275,8 @@ def before(A: T.Buffer([16, 16], "float32"), B: T.Buffer([16, 16], "float32")):

Re: [PR] [Unity][Parser] Check well-formedness in the parser [tvm]

slyubomirsky commented on code in PR #16569: URL: https://github.com/apache/tvm/pull/16569#discussion_r1505319772 ## tests/python/tir-transform/test_tir_transform_convert_ssa.py: ## @@ -327,7 +327,8 @@ class TestDeDuplicateThreadIdxAcrossMultipleFunctions(BaseBeforeAfter):

Re: [PR] [Unity][Parser] Check well-formedness in the parser [tvm]

slyubomirsky commented on code in PR #16569: URL: https://github.com/apache/tvm/pull/16569#discussion_r1505325057 ## tests/python/codegen/test_inject_ptx_ldg32.py: ## @@ -20,7 +20,8 @@ import tvm.testing -@T.prim_func +# A_local is undefined

[PR] [Web] Revert back to the non-parallel version to avoid cache.add() error [tvm]

DiegoCao opened a new pull request, #16650: URL: https://github.com/apache/tvm/pull/16650 To address the issue in https://github.com/mlc-ai/web-llm/issues/313, we revert back the parallel download to avoid the cache.error() on the webLLM side. Will dig further into the issue and come with

Re: [PR] [Unity][Parser] Check well-formedness in the parser [tvm]

slyubomirsky commented on code in PR #16569: URL: https://github.com/apache/tvm/pull/16569#discussion_r1505321744 ## tests/python/tir-transform/test_tir_transform_lower_cross_thread_reduction.py: ## @@ -116,7 +116,8 @@ def no_normal_reduction(a: T.handle, b: T.handle) -> None:

Re: [PR] [Unity][Parser] Check well-formedness in the parser [tvm]

slyubomirsky commented on code in PR #16569: URL: https://github.com/apache/tvm/pull/16569#discussion_r1505314732 ## tests/python/tir-base/test_tir_renew_defs.py: ## @@ -82,7 +82,8 @@ def _get_block(f): def test_match_buffer(): -@T.prim_func +# well-formed checker

Re: [PR] [Unity][Parser] Check well-formedness in the parser [tvm]

slyubomirsky commented on code in PR #16569: URL: https://github.com/apache/tvm/pull/16569#discussion_r1505323600 ## tests/python/tvmscript/test_tvmscript_roundtrip.py: ## @@ -27,8 +27,9 @@ def opt_gemm_normalize(): -@tvm.script.ir_module +

Re: [PR] [TVMScript] Allow use of relax.Expr with void type as a statement [tvm]

slyubomirsky commented on PR #16641: URL: https://github.com/apache/tvm/pull/16641#issuecomment-1967446140 Well, I wouldn't call that's not a type system ambiguity, but it could become one in terms of parser roundtripping. It's just that an empty tuple (unit value) cannot contain anything

Re: [PR] [Relax] Allow R.Prim('bool') in relax::If and assert_op [tvm]

slyubomirsky commented on code in PR #16642: URL: https://github.com/apache/tvm/pull/16642#discussion_r1505049818 ## python/tvm/relax/transform/transform.py: ## @@ -463,6 +463,16 @@ def KillAfterLastUse() -> tvm.ir.transform.Pass: return _ffi_api.KillAfterLastUse() #

[PR] [Frontend][PaddlePaddle] PaddlePaddle model with NHWC data format that supports quantization [tvm]

Zheng-Bicheng opened a new pull request, #16651: URL: https://github.com/apache/tvm/pull/16651 PaddlePaddle model with NHWC data format that supports quantization -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the

Re: [PR] [Relax] Allow R.Prim('bool') in relax::If and assert_op [tvm]

Lunderberg commented on code in PR #16642: URL: https://github.com/apache/tvm/pull/16642#discussion_r1505082724 ## python/tvm/relax/transform/transform.py: ## @@ -463,6 +463,16 @@ def KillAfterLastUse() -> tvm.ir.transform.Pass: return _ffi_api.KillAfterLastUse() # type:

(tvm) branch nightly updated (b3fa6cb873 -> ff3716b83a)

This is an automated email from the ASF dual-hosted git repository. github-bot pushed a change to branch nightly in repository https://gitbox.apache.org/repos/asf/tvm.git from b3fa6cb873 [AOT][Testing] Print output values on test failure (#16611) add 563ef9587c [SVE] Add support for

Re: [PR] [Unity][Parser] Check well-formedness in the parser [tvm]

Lunderberg commented on PR #16569: URL: https://github.com/apache/tvm/pull/16569#issuecomment-1967748909 I did a bisect on `test_tir_transform_force_narrow_index_to_i32.py::test_thread_axis2`, and it's been broken for quite some time. [This merge

Re: [PR] [Web] Revert back to the non-parallel version to avoid cache.add() error [tvm]

DavidGOrtega commented on PR #16650: URL: https://github.com/apache/tvm/pull/16650#issuecomment-1967881501 @DiegoCao Its not the real solution. As I stated the cache can fail even if its just one by one. I have suffered that also. Definitely HF CDN is not great. It should allow to

Re: [PR] [Web] Revert back to the non-parallel version to avoid cache.add() error [tvm]

DavidGOrtega commented on PR #16650: URL: https://github.com/apache/tvm/pull/16650#issuecomment-1967882192 I can give it a shot tomorrow -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the

Re: [PR] [Unity][Parser] Check well-formedness in the parser [tvm]

slyubomirsky commented on code in PR #16569: URL: https://github.com/apache/tvm/pull/16569#discussion_r1505315546 ## tests/python/tir-base/test_tir_specialize.py: ## @@ -65,7 +65,7 @@ def matmul_m_128(a: T.handle, b: T.handle, c: T.handle) -> None: C[vi, vj] =